الفهرس | Only 14 pages are availabe for public view |
Abstract Today, database administration is increasingly required stronger than it had ever been. Several factors have contributed in such an urgent necessity like the ever-advancing technology chiefly the data processing related tools. Most importantly, the fast growth of ever-advancing internet based applications have been the key motive behind undertaking this subject. For all these applications that are widely used on all aspects of life, Relational Database Management System (RDBMS) is still competitive as the core of the most serious production systems, and it is still more searched and managed. Moreover, such searching for managing RDBMS is originally the key mission of Database Administrators (DBA). Whether reader/user uses Oracle, Microsoft SQL Server, DB2,Informix, Sybase, MySQL, Teradata, PostgreSQL, Ingres, or any combination of these popular DBMS products, this work meant to benefit all of these application-based users. Meanwhile, Distributed Relational Database Management Systems (DDBMS) are being more than a vital necessity to acquire the results of designing Relational Database Management Systems-based application. This work then presents Distributed Database Management System (DDBMS), in the context of relational database briefly and coherently. By definition, the distributed processing is an effective way to improve the performance of RDBMS. As a result, for design RDDBMS alternatives, among the design techniques that the distributed database designer considers for performance improvement are fragmentation, replication and allocation techniques. Actually these design factors are considered as paramount research area in distributed database design environment. In one extreme, data fragmentation concerned with the scattering a relations into two or more sub-relations called fragments. Usually two fragmentation techniques are considered, these are: vertical and horizontal fragmentation. A third rarely considered one in literature which is the hybrid or mixed fragmentation. On the other extreme, the allocation of the obtained fragments to the most appropriate sites of network is still seen, in DDBS literature, a tricky task to accomplish. There is no doubt that a proper allocation for fragments across sites of network is among the most important factors that make the use for an efficient functionality of DDBS. However, the combination of fragmentation and allocation is NP-hard (and costly), as proven in literature as well, and therefore requires more perseverance to find costly-effective solutions. A heuristic approaches are nevertheless shown to be the best and cheapest ways to solve these problems by proposing some efficient settlements, specifically those of mathematically based presentation. Therefore, the work of this thesis has studies Distributed Database Management System (DDBMS) in the context of relational databases. A heuristic based fragmentation, allocation and re-allocation technique is introduced to reach the optimal or near optimal solutions for performance problems in static/dynamic DDBS environments. |